PNNL

Abstract

The Collaborative Sampling (CS) design for estimating or testing hypotheses about the mean can be more cost effective than simple random sampling under certain situations. The CS design uses two measurement methods. One method is the “standard analysis” (sometimes called the laboratory analysis or “the expensive method”). The other method is a less expensive and possibly less accurate measurement method (sometimes called the field analysis or “the inexpensive method”). The idea behind CS is to replace the need for obtaining so many expensive measurements with collecting a larger number of the less expensive measurements. Data are gathered using the inexpensive method at n’ locations. Then additional data are collected using the laboratory (expensive) method at a subset (n) of the n’ locations. CS will be more cost effective for estimating the mean than simple random sampling if there is a linear relationship with high correlation between the two types of measurements and the cost of the inexpensive method is sufficiently low. CS has recently been added to the suite of sampling designs in the Visual Sample Plan (VSP) design software, which can be downloaded free from http://dqo.pnl.gov/vsp. This presentation will outline the assumptions, methodology and possible applications of CS and present newly derived equations for computing the number of samples (n and n’) required when the sampling objective is to compute a confidence limit on the mean or test if the mean exceeds a threshold value. In addition, a real-time demonstration of the CS design as implemented in VSP will be provided.